Jan Rusz

We aimed to assess the sensitivity of key motor smartphone tasks to detect disease progression in subjects with isolated rapid eye movement sleep behavior disorder (iRBD) and early Parkinson’s disease (PD) compared to healthy controls. Twenty-six healthy controls, 22 patients with iRBD and 23 patients with early-stage PD performed speech, psychomotor, hand tremor, and U-turn tasks using a smartphone every 14 days for 24 months. Disease progression in iRBD was observed through monopitch during reading and slowing turn speed (both p < 0.05), whereas in PD, shortening of time to first voice break during phonation ( p < 0.05). Several speech, psychomotor, and U-turn measures showed adequate cross-sectional sensitivity to distinguish iRBD or PD from controls. Significant learning effect in the first year influenced measures derived from oral diadochokinesis, tapping, and hand tremor. This study represents guidance for future remote smartphone trials that integrate diverse motor domains to monitor prodromal and early Parkinsonism.

Background: The utility of high-level linguistic analysis for longitudinal monitoring of multiple sclerosis (MS) and its relationship with progressive brain volume loss remains largely unexplored. Objective: To characterize longitudinal changes in lexical and syntactic features in MS and determine whether linguistic alterations are associated with subsequent brain volume loss. Methods: Ninety-seven patients with MS and 80 age- and sex-matched healthy controls completed story narration tasks at baseline and 2-year follow-up. Recordings were automatically transcribed and analyzed to extract five linguistic features. Magnetic resonance imaging (MRI) assessments were performed between the 2- and 7-year follow-ups. Linear mixed-effects models assessed linguistic decline over 2 years and whether these early changes were associated with a subsequent 5-year brain volume loss. Results: Over 2 years, patients with MS showed changes in vocabulary range (β = 0.53, p = 0.008, q = 0.02) and phrase repetition (β = 0.56, p = 0.007, q = 0.02). Early changes in these lexical features were associated with subsequent 5-year decline in total gray matter volume. Linguistic changes were also related to cortical gray matter and cerebellar volume loss but not to hippocampal or white matter volume loss. Conclusion: Automated speech analysis captures progressive linguistic decline in MS. Early linguistic changes were associated with subsequent brain volume loss, supporting speech-derived biomarkers of neurodegeneration.

No sufficiently sensitive biomarker exists to monitor disease progression to assess treatment efficacy in synucleinopathies such as Parkinson’s disease (PD), particularly during the prodromal phase when interventions are likely to be most effective. Existing digital biomarkers often rely on active tasks or clinic-based assessments, limiting their scalability and real-world applicability. In this proof-of-concept study, we evaluated whether linguistic features derived from real-world phone call recordings using large language models can serve as a language-based progression biomarker in isolated rapid eye movement sleep behavior (iRBD). In this two-year study, we enrolled 74 participants, including 21 iRBD (20 men), 26 PD (25 men), and 27 healthy controls (26 men) age-matched to iRBD participants. Speech data collection occurred remotely in participants’ natural environments through routine phone calls. Over 34,000 phone calls (p = 0.016), semantic-syntactic diversity ( p < 0.001), topic diversity ( p < 0.001), and sentence probability ( p < 0.001) over the two years. Prodromal changes in iRBD were detectable with an area under the curve of 0.82 from as few as 21 calls. For a two-year neuroprotective trial targeting 50% drug efficacy, the estimated sample size was 78 iRBD participants per arm based on a time-to-event analysis. These findings demonstrate that fully automated phone call analysis can detect both prodromal and progressive changes in alpha-synucleinopathy. This approach is scalable, minimizes the effort required from both patients and clinical staff, and enables remote, low-burden monitoring in screening at-risk populations and therapeutic trials.

Objective: Dysarthria is one of the most common and disabling side effects of subthalamic nucleus deep brain stimulation (STN‐DBS) in Parkinson's disease (PD). Stimulation often exacerbates speech dysfunction beyond the effects of PD progression, likely because of current spread to structures surrounding the STN. This study aimed to develop speech biomarkers sensitive to DBS‐induced dysarthria by isolating stimulation side effects and mapping their emergence across incrementally increased amplitudes. Methods: Twenty‐four PD patients with bilateral STN‐DBS completed a standardized speech assessment in each hemisphere separately, including sustained phonations, rapid syllable repetitions, and reading passages across 7 increasing stimulation amplitudes defined relative to a clinically determined stimulation‐induced dysarthria threshold. A composite dysarthria index based on 7 key acoustic features, patient perceptual self‐ratings, and intelligibility scores were extracted. Results: More than 2,500 speech task recordings were analyzed. Both the composite dysarthria index and subjective self‐ratings worsened rapidly with increasing stimulation amplitude above a threshold ( p < 0.001), whereas intelligibility scores varied markedly and did not reach significance. Among individual acoustic features, phonation duration, voice quality, and monopitch exhibited significant sensitivity to increasing stimulation amplitudes. Left‐sided stimulation induced greater speech deterioration than right‐sided stimulation. Interpretation: We systematically identified speech biomarkers that capture DBS‐induced dysarthria, characterized the progressive deterioration of speech with increasing amplitude, and highlighted the pivotal role of left basal ganglia circuitry in speech production. Our objective metric holds promise as safety outcome measure for surgical therapies, guidepost for initial and troubleshooting DBS programming, and input for adaptive, closed‐loop stimulation control. ANN NEUROL 2026

Purpose: We aimed to develop a universal, fully automated segmentation algorithm that allows robust analysis of oral diadochokinesis across various neurological diseases, dysarthria types, and dysarthria severities. Method: Recordings of sequential motion rates were collected from 231 subjects, including 80 healthy controls and 151 patients with neurological diseases such as amyotrophic lateral sclerosis, essential tremor, Huntington's disease, multiple sclerosis, multiple system atrophy, Parkinson's disease, progressive supranuclear palsy, and cerebellar ataxia. A robust automatic segmentation algorithm utilizing convolutional neural networks and rule-based postprocessing was developed and evaluated across disease type, dysarthria type, and dysarthria severity. The performance of the developed artificial intelligence–based algorithm was compared with a traditional signal processing–based segmentation approach. Results: Our deep learning–based algorithm was able to correctly identify the position of individual syllables with a very high F1 score of 99.1%, compared to a signal processing–based approach with an F1 score of 97.7%. Using a 10-ms tolerance window, the deep learning–based algorithm achieved an average accuracy of 92.0% for the temporal detection of individual phoneme positions. Performance was strongly influenced by dysarthria severity, with accuracy reaching 94.7% in mild, 91.0% in moderate, and 83.1% in severe dysarthria. Disease and dysarthria type did not appear to have a substantial effect on algorithm performance. Conclusions: Our proposed deep learning–based algorithm provides reliable segmentation of syllable and individual phoneme positions during oral diadochokinesis across various disease types, dysarthria types, and dysarthria severities. The deep learning–based segmentation approaches have the potential to outperform the traditional signal processing methods for assessing oral diadochokinesis.

Assessment and monitoring of Huntington’s disease (HD) symptoms remain limited to infrequent, clinic-based evaluations. We evaluated whether fully automated linguistic analysis of speech tasks recorded via smartphone can remotely capture core clinical markers of HD severity. In this cross-sectional multicenter study across Czech and German sites, 53 participants, including 30 HD (9 pre-symptomatic, 3 prodromal, and 18 manifest) and 23 healthy controls, completed a smartphone-based speech assessment, including spontaneous monologue and fairy tale retelling for 7 consecutive days. Recordings were automatically transcribed and analyzed using natural language processing to derive 3 lexical and 3 syntactic features. Predictive models for clinical outcomes based on the Unified Huntington’s Disease Rating Scale and cognitive scales were built using multivariate linear regression with cross-validation. Linguistic features predicted HD severity across multiple domains with high predictive performance, explaining up to 57% of the variance in cognitive performance, 63% in motor impairment, and 59% in functional capacity. A median of 6 days for monologue and 3 for retelling was sufficient to reach 90% of maximal predictive performance. Compared to controls, HD participants showed reduced vocabulary range and increased phrase repetition in both tasks ( p < 0.05), with additional monologue-specific deficits in sentence length ( p = 0.018) and syntactic complexity ( p = 0.004). Fully automated analysis of smartphone-based language assessment can remotely quantify cognitive, motor, and functional impairment in HD, offering a scalable, low-burden digital biomarker for clinical trials and decentralized monitoring.

The long-term averaged spectrum (LTAS) may provide a universal method for capturing distinct patterns of dysarthria. This study aimed to evaluate the sensitivity of LTAS descriptors in a broad range of neurological diseases and various types and severities of dysarthria. Four spectral moments of spectral mean, spectral standard deviation, spectral skewness and spectral kurtosis based on LTAS were computed for reading passage collected from 461 speakers, including 306 healthy controls and 155 neurological patients secondary to Parkinson’s disease (PD), progressive supranuclear palsy, multiple system atrophy (MSA), Huntington’s disease, essential tremor, cerebellar ataxia (CA), multiple sclerosis (MS), and amyotrophic lateral sclerosis. Compared to controls, the spectral mean was significantly lower in PD and MS while elevated in CA. Significantly changed LTAS features were observed only in hypokinetic dysarthria and in mixed dysarthrias manifesting hypokinetic elements. Although LTAS features differed between controls and patients with varying degrees of dysarthria, there was no progressive increase in dysarthria severity. Our findings suggest that LTAS-based speech analysis may provide valuable cues to aid differential diagnosis among neurological diseases with overlapping clinical features. LTAS appears more informative when applied to specific diseases than to pooled dysarthria types arising from diverse neurological etiologies.

Objective: This 24‐month longitudinal study involving isolated rapid eye movement sleep behavior disorder (iRBD), early‐stage Parkinson's disease (PD), and matched healthy control subjects aimed to assess whether acoustic speech features from real‐world smartphone calls provide passive progressive biomarkers in synucleinopathies. Methods: Participants underwent clinical assessments at baseline, 1, and 2 years. Speech was continuously captured during phone calls via a standardized smartphone application, segmented, and analyzed for speech impairment severity end points and key acoustic features of monopitch, vowel articulation, voice quality, and articulation rate. We used linear mixed‐effect modeling to estimate speech progression and calculated sample size requirements to demonstrate slowing of progression under anticipated treatment effects. Results: Over 31,000 phone calls (>1,000 hours) were collected from 71 participants including those with iRBD, PD, and healthy controls. Compared with controls, both individuals with iRBD and PD showed significant declines in speech impairment severity end points based on spectral changes and artificial intelligence‐based neural embeddings. The subjects with iRBD also exhibited declines in vowel articulation and articulation rate. For a 2‐year neuroprotective trial aiming for 50% drug efficacy, the most efficient sample size estimate based on time‐to‐event analysis was 74 iRBD and 84 PD participants per arm using the neural embedding end point. Interpretation: The phone call analysis requiring no patient effort or clinical supervision can detect speech decline in prodromal and early synucleinopathies, providing a potential paradigm shift for clinical trial design and neuroprotective intervention end points. ANN NEUROL 2026;99:935–948